System Position in Global Trade Networks: Connecting Economic Inequality

Joshua Pryor
MPSA 2017 Paper Submission
NOTE: ADDITIONAL WORK STILL NEED TO BE DONE. WILL BE FINISHED AT THE
END OF SPRING BREAK 3/24/2017 AND REUPLOADED.
 In political science, network analysis has increasingly become a popular method to
analyze questions in international affairs.1 Network analysis is a method that allows for the
calculation of various indices to quantify the relational characteristics of nations. Analysts,
working in the tradition of world-systems theory, have provided empirical demonstrations
showing that a state’s position in relation to other states is seen as an important predictor of GDP
growth, political stability, and many other characteristics of states.2
Recent research has expressed a desire to unpack the relationship between world-system
position and inequality, identifying indirect mechanisms through economic development, and
direct mechanisms through “unobservables” like ruling class alliances.3 Network analysis of
global transaction flows, such as trade or foreign direct investment, provides indices to identify
where a state is located in the global economic system. Comparing network measures with
general globalization indices, such as the KOF Globalization Index, may help clarify what
network indices measure and how they are related to economic inequality.4
Bilateral data was collected for international trade using the IMF Direction of Trade
statistics database.5 N by N matrices are constructed taking 5 year averages starting from 1970
to 2009. For each matrix, several indices are calculated to understand the structure and
implications for state to state trade relationships. The major indices relevant to the network
analysis are broken into two categories. The first indicate an overall assessment of the structure
of the network, and include network centrality, network density, and a network clustering
coefficient.6 The second set of indices provides each individual country with a measure

1
See Raji Kali and Javier Reyes, “The Architecture of Globalization: a Network Approach to International
Economic Integration,” Journal of International Business Studies 38 (2007): 595-620. Matthew C. Mahutga et al.,
“Within-Country Inequality and the Modern World-System: A Theoretical Reprise and Empirical First Step,”
Journal of World-Systems Research 27 no. 2 (2011): 279-307. Michael D. Ward et al., “Network Analysis and
Political Science,” Annual Review of Political Science 14 (2011): 245-64, Emilie M. Hafner-Burton et al., “Network
Analysis for International Relations,” International Organization 63 (2009): 559-592, Stephen P. Borgatti et al.,
“Network Analysis in the Social Sciences,” Science 323 no. 5916 (2009): 892-895.
2
Foundational articles in this subject are David Snyder and Edward L. Kick, “Structural Position in the World
System and Economic Growth, 1955-1970: A Multiple Network Analysis of Transnational Interactions,” American
Journal of Sociology 84 no. 5 (1979): 1096-1126. More recent articles are Rob Clark, “World-System Mobility and
Economic Growth, 1980-2000,” Social Forces 88 no. 3 (2011): 1123-1152, and Paul Prew, “World-Economy
Centrality and Carbon Dioxide Emissions: A New Look at the Position in the Capitalist World-System and
Environmental Pollution,” Journal of World-Systems Research 16 no. 2 (2010): 162-191.
3
Mahutga et al., 2011. 281-284.
4
KOF Globalization index is from Axel Dreher, Noel Gaston, and Pim Martens, Measuring Globalization: Gauging
its Consequences (New York: Springer, 2008).
5
International Monetary Fund, Direction of Trade Statistics. Accessed December 15, 2012 at IMF.org.
6
The equations used here follow Kali and Reyes, 2007, as they use the most up-to-date network methods, in a
framework specifically tailored towards an analysis of international networks. The discussion here will not be as

2
indicating their importance and location within the network: Freeman’s degree correlation and a
core/periphery index.7 Using total network and individual country measures, the evolution of
both macro and micro levels of the global trade network is identified, providing information
about the evolution of the trade network. Subsequently, the network measures are placed within
a fixed effects regression framework to identify relationships between network measures,
globalization indices, and economic inequality.
Section one summarizes a literature review relevant to a network analysis perspective,
including the theorized connections between world-systems theory and economic inequality.
Section two is a literature of various network indices and an explanation of their purpose in
understanding the evolution of networks. Section three calculates the network indices, and then
assesses the overall structure and change of global trade networks from 1970 to 2009. In section
four, a quantitative analysis is initiated, where network indices are compared with other
measures of the international system, particularly the KOF Globalization Index, to understand
what types of relationships network indices measure. Subsequently, several models of economic
inequality are constructed to estimate the relationship between world-system position and
economic inequality.

Section 1: World-Systems Theory and Network Analysis: Making Connections with

Economic Inequality

World-systems theory was made famous by Immanuel Wallerstein, and is often seen as a
theoretical counterpart to social network analysis.8 World-systems theory argues that the
international system is characterized by three major features. First, the world economy, through
processes of globalization and integration, is becoming increasingly interconnected. Second,

detailed as in Kali and Reyes. For additional information beyond Kali and Reyes, see Wasserman and Faust, A.
Abert and A.L. Barabassi, “Statistical Mechanics of Complex Networks,” Reviews of Modern Physics 74 no. 1
(2002): 47-97, and M. Newman, “Mixing Patterns in Networks,” Physical Review 67 no. 2 (2003): 260-261, and
Newman, “The Structure and Function of ComplexNnetworks,” SIAM Review 45 no. 2 (2003): 167-256.
7
Freemans degree correlation was calculated following Paul Prew, “World-Economy Centrality and Carbon
Dioxide Emissions: A New Look at the Position in the Capitalist World-System and Environmental Pollution,”
Journal of World-Systems Research 16 no. 2 (2010): 162-191. The core/periphery index is calculated according to
Rob Clark, “World-System Mobility and Economic Growth, 1980-2000,” Social Forces 88 no. 3 (2010): 1123-1152.
8
For an in-depth theoretical review of world-systems theory and how it relates to social network analysis, see
Matthew C. Mahutga, “The Persistence of Structural Inequality? A Network Analysis of International Trade, 1965-
2000,” Social Forces 84 no. 4 (2006): 1863-1889.

3
sovereign states increasingly are the major actors that interact and are affected the world
economy. Third, the world economy is marked by a stark division of labor, where powerful
states in the ‘core’ of the world economy, use their economic and military might to extract
resources from the periphery, creating “an unequal global division of labor.”9
Core countries are powerful due to the structural advantages provided by being part of the
global core, and are characterized by diverse economies and capable of autonomous political
action. Using their privileged positions, core nations extract natural resources, goods, and other
desired material objects from a periphery of weaker nations who are unable to stop intrusions
into their societies by the core. Where the core is dominant, the periphery is weak, with
specialized economies in order to service the needs of the core, while concurrently being unable
to restrict intrusions by the core into their domestic politics. The periphery is unable to release
itself from the bonds of global servitude, as structural elements in the world system make it
difficult for a country in the periphery to climb to the core.10
While the core exploits the periphery, the semi-periphery acts as a buffer to protect the
global structural division. Including industrializing nations such as Brazil or the East Asian
tigers of South Korea and Taiwan, the semi-periphery is at once exploited by the core and
exploiter of the periphery.11 The semi-periphery is important largely because it acts as a
stabilizing feature of the world system. The semi-periphery and periphery are competitors in the
world economy, and consequently are unable to unite in order to restructure the world system in
their favor against the core, even though both the periphery and semi-periphery are exploited and
dependent on the core. If the semi-periphery ever revolts against the power of the core, system
stability disappears, and the “world system disintegrates.”12
World-systems theories have been empirically investigated using social network analysis.
Early studies demonstrated that a state’s structural position in the world system was highly
correlated with economic growth, and states within the periphery were unable to experience

9
David Knoke, Political Networks: The Structural Perspective (New York: Cambridge University Press, 1990):
183.
10
Knoke, 1990, 184. However, in a later paper, Robert Clark is able to demonstrate there are economic advantages
to moving from core to the periphery, and that movement was possible, see Robert Clark, 2010.
11
Knoke, 185.
12
Wallerstein, “The Rise and Future Demise of the World Capitalist System: Concepts for Comparative Analysis,”
in Hamza Alavi and Tedor Shanin, eds., Introduction to the Sociology of ‘Developing Societies’ (New York:
Monthly Review Press): 29-53, quoted in Knoke, 185.

4
significant economic growth.13 Analyzing eight different global transaction flows, four major
structures are identified: core, semi-core, semi-periphery, and periphery. These new sections of
the world system are theorized to increase system stability.14
Recent research using world-systems theory and network analysis have identified
structural inequalities within the international system that correspond to the core-periphery
framework. The global trade network conforms to a core-periphery framework when the
network analysis only includes country dyad pairs where trade is at least one percent of GDP.15
Snyder and Kicks early work was reexamined, and countries located within the periphery
continue to face difficulties in producing economic growth.16 Unfortunately, the core-periphery
structure has been very stable over time, meaning that countries relegated to the periphery faced
few prospects in moving to the core or the semi-periphery.17
The connection between world-system position and inequality has been frequently
studied. Position within the world system has been seen to increase economic inequality, but the
relationships and connections are still under considerable debate.18 There are several
mechanisms through which world system position can affect inequality. The first is indirect,
whereas states move closer to the core of the global system, they become more economically
developed. Following Kuznets, as a state becomes more developed, inequality begins to
increase, until eventually reaching an apex, and reducing.19 The second is a more direct
relationship, though more difficult to measure. According to world-system theorists, the primary

13
David Snyder and Edward Kick, “Structural Position in the World System and Economic Growth, 1955-1970: A
Multiple-Network Analysis of Transnational Interactions,” The American Journal of Sociology 84, no. 5 (1979):
1096-1126.
14
Edward Kick, “World-System Structure, National Development, and the Prospects for a Socialist World Order,”
in Terry Boswell and Albert Bergesen, eds., America’s Changing Role in the World-System (New York: Praeger,
1987): 127-155.
15
Raja Kali and Javier Reyes, “The Architecture of Globalization: a Network Approach to International Economic
Integration,” Journal of International Business Studies 38 (2007): 595-620,
16
The World-systems research on this particular topic consists of numerous sources. See Edward Kick and Byron
Davis, “World-System Structure and Change: An Analysis of Global Networks and Economic Growth Across Two
Time Periods,” American Behavior Scientist 44 no. 10 (2001): 1561-1571.
17
Mathew C. Mahutga, “The Persistence of Structural Inequality? A Network Analysis of International Trade, 1965-
2000,”Social Forces 84 no. 4 (2006): 1863-1889, and Rob Clark, “Dependency, Network Integration, and
Development,” Sociological Perspectives 51, no. 3 (2008): 629-648.
18
Arthur Alderson and Francois Nielsen, “Income Inequality, Development, and Dependence: A Reconsideration,”
American Sociological Review 64 (1999): 606-631. See also Linda Beer and Terry Boswell, “The Resilience of
Dependency Effects in Explaining Income Inequality in the Global Economy: a Cross-National Analysis, 1975-
1995,” Journal of world-systems Research 8 (2002): 30-59, and Cheol-Sung Lee et al, “Income Inequality, Global
Economy and the State,” Social Forces 86 (2007): 77-111.
19
Mahutga et al., 2011, 281.

5
catalyst of change in the global system is competition among capitalist economies. However, the
way that capitalist competition materializes changes over time between the core, semi-periphery,
and periphery.20 These forms of capitalist competition are often difficult to observe directly.21
Capitalist competition has different consequences depending on the location in the
international system. Core nations frequently have greater democracy as economic development
occurs, allowing for greater public participation and the formation of redistributive policies.
Periphery and semi-periphery nations, however, are often ruled by elites who are able to reduce
demands for democracy and defuse any attempts for redistribution.22
An alternative explanation connecting world-system position and inequality focuses on
the requirements for the reproduction of capitalism. Inherent within world-systems theory is a
belief that capitalism leads to an extraction of wealth from the periphery to the core. This
surplus value is consequently distributed between capitalists and workers in the core, which
decreases inequality as well as the potential for conflict. In return, inequality increases in the
periphery due to the existence of an elite capitalist class extracting surplus value from society to
give to their allies in the core. The primary catalyst in these inequality relationships is political
and economic collaboration by elites within the core and periphery to extract resources from the
periphery.23
Recent network research has supported the existence of a tightly knit web of corporate
control, possibly creating the conditions necessary for a system of exploitative elite.24 737
corporations control 80% of the economic value of over 30 million transnational corporations.
Similar to world-systems theory, a tightly connected core of 147 transnational corporations,
which have nearly full economic control of themselves, controls nearly 40% of the economic
value of all other global corporations. The overall structure of the network is reflects a core-
periphery structure, with a deeply connected core, using its wealth reserves to accumulate
significantly more wealth.25 While such a structure in itself is not proof of collusion among

20
For a more detailed explanation, see Mahutga et al., 2011, 283.
21
James Mahoney, “Revisiting General Theory in Historical Sociology,” Social Forces 83 no. 2 (2004): 459-489.
22
Mahutga et al., 2011, 283.
23
Mahutga et al., 2011, 283-284.
24
Stefania Vitali, James B. Glattfelder, and Stefano Battiston, “The Network of Global Corporate Control,” PLoS
One 6, no. 10 (2011): 1-6.
25
For a review of this phenomenon in networks, also called “The Matthew Effect,” see M. Perc, “The Matthew
Effect in Empirical Data,” The Journal of the Royal Society Interface 11, no. 98 (2014): 1-15.

6
corporate entities to extract wealth from the periphery using a web of networked elites,
cooperation among elites to reshape government policies appears plausible.

Section 2: Measuring Trade Globalization with Social Network Analysis

One of the facets that distinguish world-systems theory is its direct linkage with social
network analysis and its numerous quantitative methods to understand the structure of networks.
The major indices relevant to the current analysis are broken into two categories. The first
indicate an overall assessment of the structure of a network, and include network centrality,
network density, and the clustering coefficient.26 The second set of indices is used to place
members of a network in terms of how central or important they are in the structure of the
network, and include an individual measure of network centrality and a core/periphery measure.
This data can be used to structure the network into a core/periphery perspective and see where in
the network members are located.
Freeman’s network centrality identifies how centralized a network is as a whole. The
equation for network centrality index is:
𝑔 𝑑
∑𝑖=1[𝐶𝑚𝑎𝑥 − 𝐶𝐷 (𝑛𝑖 )]
𝐶𝐼 = 𝑔 Equation: 3.1
max ∑𝑖=1[𝐶𝑚𝑎𝑥 − 𝐶𝐷 (𝑛𝑖 )]

𝑑
𝐶𝐼 = 𝑁𝑒𝑡𝑤𝑜𝑟𝑘 𝐶𝑒𝑛𝑡𝑟𝑎𝑙𝑖𝑧𝑎𝑡𝑖𝑜𝑛 𝐼𝑛𝑑𝑒𝑥 𝐶𝑚𝑎𝑥 =
𝐴𝑐𝑡𝑢𝑎𝑙 𝑀𝑎𝑥𝑖𝑚𝑢𝑚 𝐷𝑒𝑔𝑟𝑒𝑒 𝐶𝑒𝑛𝑡𝑟𝑎𝑙𝑖𝑡𝑦 𝑂𝑏𝑠𝑒𝑟𝑣𝑒𝑑 𝑓𝑜𝑟 𝑎 𝑆𝑖𝑛𝑔𝑙𝑒 𝑁𝑜𝑑𝑒
𝐶𝑚𝑎𝑥 = 𝑇ℎ𝑒𝑜𝑟𝑒𝑡𝑖𝑐𝑎𝑙 𝑀𝑎𝑥𝑖𝑚𝑢𝑚 𝐷𝑒𝑔𝑟𝑒𝑒 𝐶𝑒𝑛𝑡𝑟𝑎𝑙𝑖𝑡𝑦 𝑓𝑜𝑟 𝑎 𝑆𝑖𝑛𝑔𝑙𝑒 𝑁𝑜𝑑𝑒
𝐶𝑑 (𝑛𝑖 ) = 𝐷𝑒𝑔𝑟𝑒𝑒 𝐶𝑒𝑛𝑡𝑟𝑎𝑙𝑖𝑡𝑦 𝑓𝑜𝑟 𝑁𝑜𝑑𝑒 𝑖

The centralization of a network quickly identifies how ‘equal’ a network is. Networks with C I
approaching one are very hierarchical and most nodes rely on a few central nodes in order to
send or receive transactions, whereas networks with low CI scores are more egalitarian and nodes
have greater access to all other nodes.27

26
The equations used here follow Kali and Reyes, 2007, as they use the most up-to-date network methods, in a
framework specifically tailored towards an analysis of international networks. The discussion here will not be as
detailed as in Kali and Reyes. For additional information beyond Kali and Reyes, see Wasserman and Faust, 1994,
Albert and Barabassi, 2002, Newman, 2003.
27
Freeman’s measure of network centrality is specifically tailored towards networks where items being transferred
pursue the shortest possible distance between nodes. As a result, this measurement is specifically tailored towards
“package delivery” processes, such as international trade. See Stephen P. Borgatti, “Centrality and Network Flow,”
Social Networks 27 (2005): 61.

7
 Network density is a measure of the total number of links between nodes divided by the
total possible number of links:
𝐿
∆𝐷 = Equation: 3.2
𝑔(𝑔−1)
∆𝐷 = 𝑁𝑒𝑡𝑤𝑜𝑟𝑘 𝐷𝑒𝑛𝑠𝑖𝑡𝑦
𝐿 = 𝑇𝑜𝑡𝑎𝑙 𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝐿𝑖𝑛𝑘𝑠 𝑖𝑛 𝑡ℎ𝑒 𝑁𝑒𝑡𝑤𝑜𝑟𝑘
𝑔(𝑔 − 1) = 𝑇𝑜𝑡𝑎𝑙 𝑃𝑜𝑠𝑠𝑖𝑏𝑙𝑒 𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝐿𝑖𝑛𝑘𝑠 𝑖𝑛 𝑡ℎ𝑒 𝑁𝑒𝑡𝑤𝑜𝑟𝑘

Network density is useful because it provides a simple measure of the level of network
integration. In this sense, a network would be fully dense if the total numbers of links that exist
are equal to the total possible number of links. As a result, it is a useful summary statistic that
briefly identifies levels of interconnectivity between members of a network.
The clustering coefficient measures the amount of multilateral engagement within an
international network. This coefficient identifies the probability that “the friend of my friend is
likely to be my friend,” and is defined as:
3×𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑇𝑟𝑖𝑎𝑛𝑔𝑙𝑒𝑠
𝐶= Equation: 3.3
𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝐶𝑜𝑛𝑛𝑒𝑐𝑡𝑒𝑑 𝑇𝑟𝑖𝑝𝑙𝑒𝑠

𝐶 = 𝐶𝑙𝑢𝑠𝑡𝑒𝑟𝑖𝑛𝑔 𝐶𝑜𝑒𝑓𝑓𝑖𝑐𝑖𝑒𝑛𝑡
𝐶𝑜𝑛𝑛𝑒𝑐𝑡𝑒𝑑 𝑇𝑟𝑖𝑝𝑙𝑒 = "𝐴 𝑆𝑖𝑛𝑔𝑙𝑒 𝑁𝑜𝑑𝑒 𝑤𝑖𝑡ℎ 𝐿𝑖𝑛𝑘𝑠 𝑅𝑢𝑛𝑛𝑖𝑛𝑔 𝑡𝑜 𝑎𝑛 𝑈𝑛𝑜𝑟𝑑𝑒𝑟𝑒𝑑 𝑝𝑎𝑖𝑟 𝑜𝑓 𝑜𝑡ℎ𝑒𝑟𝑠. "28
𝑇𝑟𝑖𝑎𝑛𝑔𝑙𝑒 = 𝑇ℎ𝑟𝑒𝑒 𝑁𝑜𝑑𝑒𝑠 𝐿𝑖𝑛𝑘𝑒𝑑 𝐷𝑖𝑟𝑒𝑐𝑡𝑙𝑦 𝑇𝑜𝑔𝑒𝑡ℎ𝑒𝑟

The clustering coefficient is useful because it acts as an indicator of the level of multilateralism
within a network.29 Triangles identify patterns of state interaction that are more dispersed
throughout the system. This indicator ranges from 0 to 1, where zero means there are no
multilateral groupings of states, whereas one means the network is perfectly multilateral, in that
all states are connected to each other through one triangle or another.
Two measures are used to identify the importance individual countries in the structure of
the network. The first measure is Prew World Centrality (Prew W.C.).30 Prew W.C. uses
Freeman’s degree centrality for each individual nation, rather than the whole network. In
creating this measure, the matrix of data is transposed on itself, creating a symmetrical matrix

28
Equation and quote taken from Kali and Reyes, 2007, 602.
29
Kali and Reyes, 2007, 602.
30
Prew, 2010, 171. I maintain the label Prew W.C. primary due to the way Prew formats the trade matrices before
calculating Freeman Centrality, and uses a natural log transformation afterwards.

8
that is intending to reduce the difference between inward and outward trade flows.31 The
calculation for the standardized country level Freeman’s centrality is:

𝑑(𝑛𝑖 )
𝐶𝐷 (𝑛𝑖 ) = Equation: 3.4
𝑔−1
𝐶𝐷 (𝑛𝑖 ) = 𝐷𝑒𝑔𝑟𝑒𝑒 𝐶𝑒𝑛𝑡𝑟𝑎𝑙𝑖𝑡𝑦 𝑜𝑓 𝑁𝑜𝑑𝑒 𝑖
𝑔 = 𝑇𝑜𝑡𝑎𝑙 𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑛𝑜𝑑𝑒𝑠 𝑖𝑛 𝑡ℎ𝑒 𝑛𝑒𝑡𝑤𝑜𝑟𝑘
𝑑(𝑛𝑖 ) = 𝐷𝑒𝑔𝑟𝑒𝑒 𝐶𝑒𝑛𝑡𝑟𝑎𝑙𝑖𝑡𝑦 𝑜𝑓 𝑎 𝑆𝑖𝑛𝑔𝑙𝑒 𝑁𝑜𝑑𝑒

Following Prew, the calculations of Freeman’s centrality for each individual country uses raw
bilateral country data. The natural log of Freeman’s degree centrality is used to reduce skew.32
The second measure of world system position is the continuous core/periphery
coefficient, which is calculated for each country:33
𝜌 = ∑𝑖,𝑗 𝑎𝑖𝑗 𝛿𝑖𝑗 Equation: 3.5
𝛿𝑖𝑗 = 𝑐𝑖 𝑐𝑗
𝜌 = 𝑆𝑖𝑚𝑖𝑙𝑎𝑟𝑖𝑡𝑦 𝑜𝑓 𝑎 𝑛𝑒𝑡𝑤𝑜𝑟𝑘 𝑡𝑜 𝑎 𝑐𝑜𝑟𝑒/𝑝𝑒𝑟𝑖𝑝ℎ𝑒𝑟𝑦 𝑠𝑡𝑟𝑢𝑐𝑡𝑢𝑟𝑒
𝑎𝑖𝑗 = 𝑃𝑟𝑒𝑠𝑒𝑛𝑐𝑒 𝑜𝑓 𝑎 𝑙𝑖𝑛𝑘 𝑏𝑒𝑡𝑤𝑒𝑒𝑛 𝑛𝑜𝑑𝑒𝑠 𝑖 𝑎𝑛𝑑 𝑗
𝑐 = "𝐴 𝑣𝑒𝑐𝑡𝑜𝑟 𝑜𝑓 𝑛𝑜𝑛𝑛𝑒𝑔𝑎𝑡𝑖𝑣𝑒 𝑣𝑎𝑙𝑢𝑒𝑠 𝑖𝑛𝑑𝑖𝑐𝑎𝑡𝑖𝑛𝑔 𝑡ℎ𝑒 𝑑𝑒𝑔𝑟𝑒𝑒 𝑜𝑓 𝑐𝑜𝑟𝑒𝑛𝑒𝑠𝑠 𝑜𝑓 𝑒𝑎𝑐ℎ 𝑛𝑜𝑑𝑒"34

Table 1.1 identifies an ideal core/periphery network. Actors one through four represent
the core, because they are connected to each other (a connection is represented by a one, no
connection is represented by a zero), as well as to nodes five through ten, which consist of the
periphery. Since the periphery is not connected to each other, in order for one node in the
periphery to contact another periphery node, they must go through the core. This ‘gatekeeper’
power that the core acquires demonstrates the dependency of the periphery on the core.35 While
real-world networks rarely perfectly mimic the idealized model, the closer the real world

31
Prew, 2010, 171. For additional information on the transposing to reduce the impact of inward and outward types
of transactions, see Tieting Su, “Myth and Mystery of Globalization: World Trade Networks in 1928, 1938, 1960,
and 1999,” Review 25 no. 4 (2002): 351-392. During direct correspondence with Prew, he clarifies this point further,
arguing that the best way to capture a nation’s role in the international system is taking into account inflows and
outflows into the overall system. This correspondence can be provided on request, and is dated July 23, 2012.
32
Prew, 2010, 173.
33
Discussion of the core/periphery measure is in Borgatti and Everett, 1999. The use of the continuous
core/periphery measure in the analysis of the international system is implemented in Rob Clark, 2009.
34
Borgatti and Everett, 1999, 387.
35
Historical examples of gatekeepers in social networks is in Mamadi Corra and David Willer, “The Gatekeeper,”
Sociological Theory 20, no. 2 (2002): 180-207. Gatekeepers have high levels of “social capital,” see Pierre
Bourdieu, “The Forms of Capital,” in John G. Richardson, ed., Handbook of Theory and Research for the Sociology
of Education, (New York: Greenwood Press, 1986): 348.

9
network is to this idealized form, the greater the opportunities for exploitation and increasing
likelihood for inequality.

Table 1.1: Idealized Core/Periphery Network

Source: Borgatti and Everett, 1999, 378.
1 = Existence of a connection between actors
0 = Lack of a connection between actors
Actors
Core Periphery
1 2 3 4 5 6 7 8 9 10
1 1 1 1 1 1 1 1 1 1
Core

2 1 1 1 1 1 1 1 1 1
3 1 1 1 1 1 1 1 1
4 1 1 1 1 1 1 1 1 1
Actors

5 1 1 1 1 0 0 0 0 0
6 1 1 1 1 0 0 0 0 0
Periphery

7 1 1 1 1 0 0 0 0 0
8 1 1 1 1 0 0 0 0 0
9 1 1 1 1 0 0 0 0 0
10 1 1 1 1 0 0 0 0 0
The core is connected with all other actors, whereas the periphery is only connected to the
core, and not to other periphery actors. A core-periphery structure helps the core to determine
access to other actors throughout the network.

In calculating the core-periphery measure, following Clark (2011), the value of trade
between each individual country is base 10 logged to reduce skew.36 The core-periphery
measure is then calculated for a matrix as well as the transposition of the matrix, in order to
measure both inflows and outflows of trade. The resulting core-periphery measures are then
normalized with a mean of zero and a standard deviation of one, and are subsequently averaged
to get a final core-periphery measure. There may be a slight problem with this procedure, as the
core-periphery routine works best when the data matrix is symmetrical.37 As a result, log base 10
trade matrices are transposed and added on themselves to create a symmetrical trade matrix that
includes both imports and exports in each countries measures, and core-periphery is then
calculated. Both types of core-periphery measures are used in the following analysis.

36
Clark, 2010, 1130.
37
Personal correspondence with Martin Everett, one of the developers of the core-periphery measure.
Correspondence is available upon request.

10
 Prew (2010) argues that the core/periphery index is a misidentified measure of world
system position, even though core-periphery has been used frequently in world-systems
literature.38 The core/periphery measure weights the importance of an individual country with
the coreness of their partners. Prew argues this creates bias because, if nations that are in the
periphery trade heavily with core nations, then those periphery nations can have an inflated
score. With raw trade data, for example, Canada and Mexico are located extremely close to the
core of the trade network because of their high trade volume with the United States, even though
Mexico is not typically seen as a member of the global economic core.39 However, Clark (2011)
does not use raw trade data when calculating the core-periphery measure, and instead takes the
base 10 logarithm of trade values, leading to measures of world system position that reflects
world-system theory better than the raw data. To identify which measure of world system
position is more useful in longitudinal time series analysis, both the Prew W.C. and Clark’s core-
periphery measure are used.

Section 3: Summary of Network Analysis Statistics

To understand how global trade networks have changed, trade flows for a sample of 137
countries is taken from 1970 to 2009.40 Data between nations is averaged for five year periods,
creating eight trade networks for 1970-74, 1975-79, 1980-84, 1985-89, 1990-94, 1995-99, 2000-
04, and 2005-09. Following Kali and Reyes, overall network measures are presented for each
network in Table 1.2.41 The measurements include network centralization, network density, and
the clustering coefficient. For these calculations, the trade networks are dichotomized, so there
is a 0 for no relationship, and 1 when there is trade between nations. In Kali and Reyes’s
original analysis, they calculated these overall statistics for different thresholds, at 0% (all trade),
1% (trade between nations is at least 1% of a nation’s total trade), and 2% (trade between nations
is at least 2% of a nation’s overall trade). 42 As the threshold increases, the matrices become

38
Prew, 2010, 171-172. For instances of the core-periphery measure in the world system literature, see Robert
Clark, 2010, and Mahutga et al., 2011.
39
Authors calculations and also Prew, 2010, 172.
40
137 countries are chosen because data is available for them during the whole period of the analysis in the IMF
Direction of Trade Statistics. Countries will be listed in the appendix (not finished yet, but immediately available
upon request).
41
Kali and Reyes, 2007, 601.
42
Kali and Reyes, 2007, 598.

11
more centralized and less dense, as is expected given the analysis is focusing more on trade-
dependent nations. These cutoffs are not used in the current analysis because the focus is on the
overall image of international trade.

Table 1.2: Network Statistics Overview– 1970 to 2009

1970- 1975- 1980- 1985- 1990- 1995- 2000- 2005-
74 79 84 89 94 99 04 09
Centrality
(Indegree) 41.863 37.905 42.46 38.286 30.713 24.118 21.736 21.399
Density 0.506 0.516 0.582 0.623 0.697 0.762 0.786 0.789
Clustering
Coefficient 0.677 0.674 0.741 0.765 0.804 0.834 0.849 0.854

Table 1.2 is a summary of important network measures from 1970 to 2009. Except for a
slight increase in the 1980-84 period, the networks have become less centralized. This indicates
that individual states are more connected to each other, and the influence of a core of powerful
trading nations is reduced. Overall, network density is increasing, which means that the number
of actual connections between states compared to the total possible number of connections is
increasing, indicating that states have more and more trade relationships and the global trade
network is becoming increasingly integrated. A similar phenomenon is seen with the clustering
coefficient, which identifies whether or not the trading partner of one nation is also the same
partner of another. The clustering coefficient is increasing throughout the time analyzed, again
demonstrating that the global trade network is increasingly integrated and connected.
However, these results should be interpreted in the light that it includes all the raw trade
data between nations, even if the level of trade is minimal between two states. If the analysis
focused only on trade amounts that were a higher percentage of a state’s overall trade, then as
seen in Kali and Reyes, the global trade network would be much more centralized. Regardless,
with the raw data, the story of global trade relationships is one of an increasingly connected and
globalized world, where states have greater opportunities to find new trading partners.
Another way to visualize the world trading system is by translating core-periphery scores
into Euclidean space. To give an idea of where countries are located, Table 3.2 provides the top
and bottom ten countries for 1970-74 and 2005-09. The core-periphery data derived from the
symmetric data is used to create the visualization. For comparison, the Prew W.C. scores are
also provided for the same years.

12
 Graph 1.1 and Graph 1.2 provide a visualization of the world trading system for 1970-74
and 2005-09 respectively. Nations colored red are in the core, and nations colored blue are in the
periphery. Note that the core of the graphs is clustered and not all the names are visible. The
country names from Table 3.2 help identify the countries located in the core. As seen in Table
3.2 and supported as well by Graph 1.1 and 1.2, there are large amounts of change and stability
within the global trading network. China’s rise as a global trading powerhouse is clearly visible.
In 1970-74, China is in the periphery, located in the north of Graph 1.1, whereas in Graph 1.2
China is located in the middle of the core and cannot be seen on the graph, though it is ranked
second on Table 1.3 for 2005-09. These visualizations demonstrate the feasibility of movement
from the periphery to the core of the world-system. Clark (2011) shows from 1980 to 2000 that
movement from the periphery to the core was not only feasible, but a significant catalyst for
economic growth. The data from the current data supports Clark’s analysis, which argues
against other analysis such as Mahutga (2011), which maintains that a state’s position within the
world-system is fairly constant.43

43
Mahutga, 2011, 291, where they are analyzing the significance of world-system position “under the assumption
that world-system position is a more or less time invariant (or nearly invariant) attribute that varies across
countries.”

13
 Table 1.3: Top 10 and Bottom 10 Rankings for Different World Centrality Measures 1970-74 and 2005-09
Country C-P Sym Country C-P Sym Country Prew W.C. Country Prew W.C.
1970-74 2005-09 1970-74 2005-09
United States 2.84 United States 1.72 United States 25.53 United States 28.67
UK 2.81 China 1.65 Germany 25.18 Germany 28.24
Germany 2.81 Germany 1.62 Japan 24.76 China 28.24
France 2.68 France 1.52 UK 24.73 Japan 27.80
Japan 2.68 Italy 1.48 France 24.66 France 27.58
Italy 2.58 UK 1.48 Canada 24.55 UK 27.48
Netherlands 2.48 Japan 1.48 Italy 24.41 Italy 27.40
Switzerland 2.15 Netherlands 1.45 Netherlands 24.24 Netherlands 27.36
Spain 2.05 India 1.38 Sweden 23.65 Canada 27.35
Sweden 2.05 Spain 1.38 Switzerland 23.56 Korea, South 27.21
~~~~~ ~~~ ~~~~~ ~~~ ~~~~~ ~~~ ~~~~~ ~~~
Mauritania -1.70 Chad -1.07 Rwanda 17.60 Rwanda 20.53
Grenada -1.77 EG -1.07 Guin-Bis 17.54 Gambia 20.46
Albania -2.00 Laos -1.10 EG 17.52 Sierra Leone 20.37
Comoros -2.10 Somalia -1.17 Albania 17.39 Cape Verde 20.31
EG -2.13 Fiji -1.17 Nepal 17.09 Guin-Bis 19.79
STP -2.17 Burundi -1.24 Cape Verde 17.07 Grenada 19.75
Cape Verde -2.17 Grenada -1.34 Comoros 16.84 Burundi 19.64
Mongolia -2.20 Comoros -1.37 Grenada 16.80 CAR 19.62
Nepal -2.23 Guin-Bis -1.44 STP 16.55 Comoros 19.02
Zimbabwe -2.30 STP -1.57 Mongolia 16.20 STP 18.31
EG: Equatorial Guinea Guin-Bis: Guinea-Bissau STP: Sao Tome and Principe UK: United Kingdom

14
Graph 1.1: Core and Periphery in the World Trading System: 1970-74
Red = Periphery
Blue = Core

15
Graph 1.2: Core and Periphery in the World Trading System: 2005-09
Red = Periphery
Blue = Core

16
 Section 4: Quantitative Analysis of Network Measures

What do network indices of world trade networks actually measure? To further advance
an understanding of networks indices, they are compared with the KOF index of globalization,
and various inequality measures.44 Table 1.4 is a correlation matrix of the KOF globalization,
economic inequality, and network measures. The first observation of importance is how the
three network indices are correlated the least with the economic measure of globalization. The
economic measure consists primarily of flows between states, such as trade and FDI, as well as
restrictions, such as tariffs. The network indices are all based on bilateral trade flows, meaning
they should be highly correlated with a measure of globalization that includes trade flows. Since
the network measures are most correlated with political globalization, and least correlated with
economic globalization, this encourages a view where system position, as measured by network
indices of trade flows, reflects more than just the economic connections of states. States are able
to achieve highly central positions within networks not simply because of the level of trade
flows, but likely due to participation in global institutions, or the use of political or military
force.
In regards to inequality, there is a sharp distinction between pre- and post-tax
redistribution Gini inequalities and the network measures. Pre-tax redistribution inequality has
very little correlation with the network measures, but a medium negative correlation occurs with
after-tax redistribution. This relationship is observed for all network indices, perhaps indicating
that the more centrally located a nation is in the global trade network, the greater the influence
tax redistribution has, or the greater the capability a state has in using the tax code to ameliorate
the effects of globalization.
The last intriguing observation is that the Clark C-P and the Symmetrical C-P measures
are practically the same. Clark (2011) makes a strong argument for his two-step procedure,
though he doesn’t address the symmetrical alternative.45 This correlation matrix indicates that
the final product from both methods is approximately the same.

44
The format for the following section is influenced by Mahutga et al., 2011, 279-307. The KOF index is from
http://globalization.kof.ethz.ch/, accessed July 30, 2013. Wage inequality is from
http://utip.gov.utexas.edu/data.html, accessed July 30, 2013. Pre- and post-tax redistribution inequality is from the
Standardized World Income Inequality Database at http://myweb.uiowa.edu/fsolt/swiid/swiid/updates.html, accessed
July 30, 2013. Other measures are noted as necessary.
45
Clark, 2009, 1130-1131.

17
 Table 1.4: Correlations Between Globalization, Inequality, Political Stability, and Network Measures

Econ. Soc. Pol. Gini Gini Post- Wage Clark Sym. Prew
Glob Glob Glob Pre-Tax Tax Inequality. C-P C-P W.S.
Econ. Glob 1
Soc. Glob 0.82 1
Pol. Glob 0.42 0.4 1
Gini - Pre
Tax -0.28 -0.46 -0.31 1
Gini-Post
Tax 0 -0.08 -0.15 0.78 1
Wage Ineq. -0.23 -0.34 -0.32 0.62 0.4 1
Clark C-P 0.44 0.58 0.74 -0.48 -0.2 -0.39 1
Sym. C-P 0.43 0.57 0.73 -0.48 -0.2 -0.4 1 1
Prew W.S. 0.58 0.63 0.77 -0.42 -0.21 -0.32 0.87 0.86 1

18
 The first series of regressions seeks to compare the network indices with the KOF
globalization measures.46 The question is whether or not the KOF and network indices measure
similar facets of the international system, or entirely different ones. A fixed effects model with
cluster robust standard errors and time dummies reduces the impact of heteroskedasticity and
serial correlation. The results of the regressions are in Table 1.5. Economic and social
globalization, which primarily reflects direct ties between nations, is significantly related with
Prew W.C. The Prew measure takes the natural log of the total amount of trade into and out of a
nation, making it similar to the KOF economic and social measures because they are all
measuring the strength of ties.
The core-periphery measures are insignificant for economic and social globalization. The
core-periphery measures, rather than focusing on the strength of direct ties between countries,
instead measures how far each country is from the predicted center of the network based on the
hypothetical core-periphery model. Rather than focusing on ties between countries, core-
periphery measures each countries distance from the core, with higher values indicating a closer
proximity to the core. This measure may experience significant problems if the network being
analyzed is not structured similar to the core-periphery hypothesis. What makes a core-periphery
framework unique is how the core is highly integrated with the core as well as the periphery, but
the periphery is also comparatively less connected with itself. This provides the core a type of
“network power” which the core can use to structural relations among actors to the core’s
benefit. The overall network measures discussed previously identify a global trade network that
is increasingly connected and integrated, between all parts, including both a theoretical core and
periphery. Hence, the insignificance of the core-periphery measures in these regressions may
reflect the highly interconnected nature of the global trading system, and that when all levels of
trade are included, the system does not reflect a core-periphery structure.

46
This section is directly influenced by Mahutga et al., 2011, starting at page 290.

19
 Table 1.5: Globalization and World Position Indices
Fixed Effects with Time Dummies and Cluster Robust S.E
5 Year Avg. Data
(1) (2) (3) (4) (5) (6) (7) (8) (9)
VARIABLES Econ. Econ. Econ. Soc. Soc. Soc. Pol. Pol. Glob. Pol.
Glob. Glob. Glob. Glob Glob Glob. Glob. Glob.

Clark C-P 0.133 -1.464 3.830***

(0.821) (0.941) (1.092)
Sym. C-P -0.00619 -1.527* 3.595***
(0.791) (0.900) (1.032)
Prew W.C. 2.421*** 2.434*** 0.978
(LN) (0.616) (0.734) (0.678)
Constant 38.75*** 38.75*** -11.75 31.10*** 31.13*** -19.18 42.32*** 42.23*** 21.81
(0.489) (0.489) (12.87) (0.586) (0.579) (15.32) (0.641) (0.638) (14.06)

Time Dummies Yes Yes Yes Yes Yes Yes Yes Yes Yes
Observations 920 920 920 1,037 1,037 1,037 1,053 1,053 1,053
R-squared 0.723 0.723 0.737 0.604 0.605 0.623 0.709 0.709 0.700
# of Countries 117 117 117 132 132 132 134 134 134
Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1

20
 Intriguingly, the core-periphery index is highly significant with the KOF measure for
political globalization. One possible interpretation of this relationship is an affirmation of the
power relations structured by a core-periphery framework. The relationship is positive,
indicating that countries located closer to the core of the world-system are more connected in
political relations, such as treaties and membership in international organizations. These
political connections allow core states greater influence in structuring the international system in
ways that benefits their own interests. Unfortunately, the relationship is not extremely powerful.
The r2 in regressions 7 and 8 in table 1.5 is .7 primarily due to the influence of the time dummies.
When the model is run without them, the within r2 is reduced to .07, although the relationship is
still highly significant. If the core-periphery measure does identify some form of political
relationship between states, it is a weak measure; i.e., there is lots of noise in the statistic.
To extend the understanding of what these network measures actually measure, several
institutional variables are used as independent variables. To identify whether or not these
network indicators proxy institutional development, data is gathered from the Economic
Freedom of the World project.47 Three measures are chosen to reflect institutional development:
size of government, legal system and property rights, and regulation. From the Penn World
Tables, GDP per Capita is also included to reflect economic development.48

47
www.freetheworld.com
48
For a brief review on elements of institutional strength, see Steven Levitsky and Maria Victoria Murillo,
“Variation in Institutional Strength,” Annual Review of Political Science 12 (2009): 115-33. On the quantitative
measurement of institutional quality, see Fabrizio Carmignani, “The Distributive Effects of Institutional Quality
when Government Stability is Endogenous,” European Journal of Political Economy 25 (2009): 409-421.

21
 Table 1.6: Network Indices and Institutional Development
Fixed Effects Regressions with Time Dummies and Cluster Robust S.E.
(1) (2) (3)
VARIABLES Clark C-P Sym. C-P Prew W.S.

Gov’t Size 0.0174 0.0186 -0.0254*

(0.0192) (0.0197) (0.0150)
Legal System -0.00273 -0.00327 -0.0276*
(0.0167) (0.0172) (0.0144)
Regulation 0.0161 0.0163 0.0901***
(0.0227) (0.0232) (0.0240)
GDP Per Capita (LN) 0.170 0.179 1.008***
(0.151) (0.156) (0.104)
Tech Development 0.175*** 0.178*** 0.144***
(LN Trademarks) (0.0564) (0.0576) (0.0433)
Constant -2.186 -2.216 11.84***
(1.338) (1.383) (0.871)

Observations 510 510 510

R-squared 0.155 0.181 0.958
Number of Countries 92 92 92
Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1

Including measures of institutional development identifies a clear divide between the

core-periphery and the Prew W.S. indicators, as seen in Table 1.6. Prew W.C. measures the total
value of imports and exports to a country and is strongly related with regulation and levels of
economic development, indicating that institutions develop additional forms of control in
response to greater regulatory demands.49 The greater the amount of transactions, the more
regulations required to manage those transactions. The level of economic development
theoretically should also be significant because as economic development increases, more
transactions should occur, and states should achieve a greater level of centrality, which would be
picked up in the Prew W.S. indicator. The insignificance of government size and the legal
system indicate that higher levels of centrality have a targeted relationship with a country’s
regulatory infrastructure, but not so much with other parts of government.

49
This finding supports transaction analysis theory, where transactions between states creates pressures to develop
institutions. See Karl W. Deutsch, “Power and Communication in International Society,” in A.V.S. deReuck and
Julie Knight, eds., Conflict in Society (London: K. & A. Churchill, Ltd., 1966): 300-316.

22
 The core-periphery indicators did not have significant relationships with most of the
indicators of government development. This could imply several possibilities. The first is that,
as previously discussed by Mahutga, the core-periphery measures are able to capture the
unobservable, or difficulty to quantify struggles that occur as a result of the capitalist economic
system, where elites struggle for power and control in the international system. Unfortunately, it
seems rather difficult to empirically demonstrate this type of relationship. Alternatively, the
international system could simply not reflect a core-periphery structure, and it is miss-specified
in the analysis of the world-system.
Levels of technological development are significant in explaining both core-periphery
and Prew measures. For core-periphery, countries with a higher level of technological
development are closer to the center of the global trading network, as well as having a higher
position in the world-system. Technology helps increase the total flow of transactions, while
concurrently helping nations move closer to the core of the global trade network.
The next series of regressions tries to establish a relationship between inequality and
network measures. Two mechanisms relate inequality and world-systems measures. First, world
system processes, by affecting economic development, can affect inequality through economic
development and the Kuznets model. Alternatively, world-systems dynamics can affect
inequality directly, by affecting capitalist modes of production and economic elites.50 Empirical
connections between world-systems theory and inequality have been tenacious, in part due to the
difficulty of operationalizing underlying mechanisms or perhaps flaws in the overall world-
systems theory. Previous literature that attempts to find a direct relationship between world
system position and inequality typically find a negative relationship, where more core nations
have less inequality, however this relationship disappears when controlled for economic or
institutional development.51
The results of several fixed effects models between world-system indicators and
economic inequality are displayed in Table 1.7. Before adding in additional controls, these
regressions seek to identify a direct relationship mimicking the “unobservables” discussed by
Muhatga (2011). In all models, the core-periphery measure is never significant, reinforcing the

50
Mahutga et al., 2011, has a larger summary of previous research in this area.
51
Cheol-Sung Lee, “Income Inequality, Democracy, and Public Sector Size,” American Sociological Review 70
(2005): 158-181, and Cheol-Sung Lee et al., “Income Inequality, Global Economy and the State,” Social Forces 86
(2007): 77-111.

23
difficulties others have observed in finding his type of link. Intriguingly though, Prew W.S is
highly significant for all types of inequality. Prew W.S. has a smaller coefficient in the after-tax
redistribution Gini than the pre-tax gini, perhaps showing how more developed government
institutions are better at reducing inequality. However, the coefficient is still positive in both
cases, demonstrating that as a nation becomes more central to the world trade network as
measured by the absolute amount of trade flows, the greater the amount of inequality.
Alternatively, wage inequality has a negative coefficient, meaning that as a nation has greater
inflows and outflows of trade, the less wage inequality they experience.
Table 1.8 analyzes the network-inequality relationship with the inclusion of a basic
development model. Inequality is explained as the result of Prew W.S., GDP per Capita,
secondary school enrollment, and the rate of population growth. Prew W.S. is found to have a
statistically significant positive increase on pre-tax inequality, even with the additional controls.
However, the effect of Prew W.S. on after-tax redistribution inequality is insignificant. The
additional controls also remove the previously seen relationship between wage inequality and
Prew W.S. This indicates that whatever the impact the Prew W.S. measure of world-system
position has on economic inequality, state institutions are able to placate those affects through
the use of the tax redistribution system. As a result, this analysis encourages a view where the
state has a strong role to play in determining the effect of continuing integration with the world-
system.

24
 Table 1.7: Inequality Measures and Network Indices
Fixed Effects with Time Dummies and Cluster Robust S.E.
(1) (2) (3) (4) (5) (6) (7) (8) (9)
VARIABLES Pre-Tax Pre-Tax Pre-Tax After-Tax After-Tax After-Tax Wage Wage Wage
Gini Gini Gini Gini Gini Gini Ineq Ineq Ineq

Clark C-P -1.065 0.227 0.0547

(1.097) (0.779) (0.0424)
Sym. C-P -1.062 0.216 0.0527
(1.041) (0.738) (0.0404)
Prew W.S. 2.776*** 1.326** -0.0753**
(0.706) (0.619) (0.0317)
Constant 48.58*** 48.63*** -10.65 38.87*** 38.86*** 10.80 -1.482*** -1.484*** 0.145
(0.947) (0.968) (15.09) (0.677) (0.692) (13.20) (0.0265) (0.0271) (0.675)

Observations 709 709 709 714 714 714 592 592 592
R-squared 0.076 0.076 0.120 0.019 0.019 0.040 0.233 0.233 0.240
# of Countries 121 121 121 121 121 121 115 115 115
Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1

25
 Table 1.8: Prew W.S. and Inequality with Control Variables
Fixed Effects with Cluster Robust S.E.
(1) (2) (3)
VARIABLES Pre-Tax Post-Tax Wage
Inequality Inequality Inequality

Prew W.S. 2.034** 1.097 0.0253

(0.894) (0.983) (0.0465)
GDP Per Capita (LN) 3.049 0.643 -0.186*
(2.265) (1.592) (0.106)
Secondary School Enrollment -0.0360 -0.0363 -0.00519**
(0.0372) (0.0268) (0.00226)
Rate of Population Growth 0.341*** 0.100 -0.00451
(0.126) (0.0868) (0.00587)
Constant -25.52 9.709 -0.140
(20.72) (16.43) (0.736)

Time Dummies Yes Yes Yes

Observations 650 655 548
R-squared 0.177 0.046 0.273
Number of Countries 118 118 110
Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1

The previous world system position – inequality regressions try to identify a direct link
through unobservables and the inherent characteristics of the network measurements. Muhutga
argues there is an alternative mechanism through which world system position can affect
inequality: by encouraging economic growth. Recently, Clark (2011) tested the world system
position – economic growth relationship through a fixed effects model. Using the variables from
Clark, the following fixed effects model is tested:

Economic Development (LN GDP Per Capita PPP)it =

βGross Capital Formation (% GDP))it +
βSecondary School Enrollmentit +
βLabor Force Participation (% Total Population) it +
βWorld-System Positionit +
βctrycodeit +
βyearit +
εit

where i is country and t is time. This model attempts to re-create a neoclassical economic
growth model. Growth is the result of capital inputs (gross capital formation) and labor

26
(secondary school enrollment and labor force participation).52 The results of the models are in
table 1.9.
The models clearly demonstrate the importance of world system position in positively
influencing economic growth. The core-periphery style measures were significant and positive
at the .05 level, and the Prew W.S. centrality measure was positive and significant at the .001
level. These results closely mimic those found in Clark (2011). The more central and the closer
to the core of the global trade network, the more economic growth a nation is likely to have. To
the extent that economic growth influences inequality, world system position indirectly affects
inequality through the classic Kuznets model.

Table 1.9: World System Position and Economic Growth

Fixed Effects Regressions with Time Dummies and Cluster Robust S.E.
(1) (2) (3)
VARIABLES GDP Per GDP Per GDP Per
Capita (LN) Capita (LN) Capita (LN)

Capital Formation 0.00517* 0.00514* -0.00107

(% GDP) (0.00266) (0.00265) (0.00149)
Secondary School. -0.00166 -0.00167 -0.00237
Enrollment (0.00208) (0.00208) (0.00161)
Labor Force -0.00162 -0.00149 0.00202
Participation (0.00586) (0.00585) (0.00358)

Clark C-P 0.167**

(0.0834)
Sym. C-P 0.165**
(0.0805)
Prew W.S. 0.389***
(0.0845)
Constant 8.502*** 8.496*** -0.395
(0.415) (0.414) (1.891)

Observations 451 451 451

R-squared 0.517 0.517 0.728
Number of ccode 125 125 125
Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1

52
For more details, see Clark, 2010, 1131-1132.

27
 Section 5: Conclusion

A network analysis of the world trade system provided numerous observations about the
structure and behavior of the world system. Network wide measurements indicated that the
global trade network is becoming more and more integrated, with the “core” increasing in size,
and more and more states having direct connections with each other. The world system position
indicators that were calculated, both the core-periphery measures and the centrality measure, had
significant explanatory power in several regression frameworks.
Prew (2009) argued that centrality measures are more adequate in the analysis of
networks than core-periphery networks. His argument was based on how the core-periphery
index was calculated. The core-periphery index, by weighting an individual state’s core-
periphery score based on how strong the connections of that state’s direct partners were, created
bias in perceptions of the global network. Regardless of the statistical weighting issues of the
core-periphery indicators, the primary problem with the measure is that it forces global relations
into a core-periphery structure, even if that structure does not accurately describe the world
system. For a network to be a core-periphery structure, the core has many connections with itself
and the periphery, whereas the periphery has most of their connections with just the core, and not
other periphery nations. As indicated by the overall network indicators, network centrality,
network density, and the clustering coefficient, the overall trade network is becoming
increasingly integrated, including core-periphery, and most importantly, periphery-periphery
relations. The power of the core comes from the relative isolation of states located within the
periphery; the main power of the core comes from their gatekeeping power, where peripheral
states need to go through the core to gain access to the rest of the network. That pattern appears
less significant than previously articulated. While the global trading system during
colonialization was a highly centralized core-periphery network, the current analysis of the last
40 years indicates that the world system is becoming less hierarchical and more peer-to-peer.
These changes in the global trade network encourage cautious usage of core-periphery
measurements in the future. Indeed, in most of the models demonstrated here, the Prew W.S.
measure often had more significance and larger coefficients than the core-periphery measures.
The key difference is that the Prew measure does not force the data into a pre-conceived model
of what the network looks like, rather it just identifies the total sum of all the connections a state

28
has with the rest of the network. This particular measure proved extremely robust in all the
analysis pursued here.
However, the core-periphery measures were able to capture an element of the world
system that the Prew measure could not, although perhaps weakly: structural power. In the
globalization regressions, the core-periphery measures were highly significant with political
globalization structures, whereas the Prew measures were more related with transactional
globalization measures. The core-periphery indices may provide avenues to examine the
“unobservables,” such as elite manipulation of capitalist modes of production that seems to
escape the Prew measures. Future work should use other inter-state transactions, such as foreign
direct investment, to see if a core-periphery structure is applicable.

29
 Bibliography

Albert, A. and A.L. Barabassi. “Statistical Mechanics of Complex Networks.” Reviews of

Modern Physics 74 no. 1 (2002): 47-97.

Alderson, Arthur S. and Francois Nielsen. “Income Inequality, Development, and Dependence:
A Reconsideration.” American Sociological Review 64 no. 4 (1999): 606-631.

Beer, Linda and Terry Boswell. “The Resilience of Dependency Effects in Explaining Income
Inequality in the Global Economy: a Cross-National Analysis, 1975-1995.” Journal of World-
Systems Research 8 (2002): 30-59.

Borgatti, Stephen P. and Martin G. Everett. “Models of Core/Periphery Structures.” Social

Networks 21 (1999): 375-395.

Borgatti, Stephen P. “Centrality and Network Flow.” Social Networks 27 (2005): 55-71.

Borgatti, Stephen P. “Network Analysis in the Social Sciences.” Science 323 no. 5916 (2009):
892-895.

Boswell, Terry and Albert Bergesen, eds. America’s Changing Role in the World-System. New
York: Praeger, 1987.

Richardson, John G. Handbook of Theory and Research for the Sociology of Education. New
York: Greenwood Press, 1986.

Carmignani, Fabrizio. “The Distributive Effects of Institutional Quality When Government

Stability is Endogenous.” European Journal of Political Economy 25 (2009): 409-421.

Chadwick, Richard and Karl Deutsch. “International Trade and Economic Integration: Further
Developments in Trade Matrix Analysis.” Comparative Political Studies 6 no.1 (1973): 84-109.

Chadwick, Richard W. “Reframing the Meaning of Democracy: The Globalization of

Democratic Development, Viewed through the Paradigms of Political Science, Political Practice,
and Political Philosophy.” Paper presented to the International Political Science Association
Conference in Fukuoka, 2006. Available online at
http://www.hawaii.edu/intlrel/fukuoka2006_paper_97.pdf.

Clark, Cal, and Susan Welch. “Western European Trade as a Measure of Integration: Untangling
the Interpretations.” Journal of Conflict Resolution 16 no. 3 (1972): 363-382.

Clark, Rob. “Dependency, Network Integration, and Development.” Sociological Perspectives

51, no. 3 (2008): 629-648.

Clark, Rob. “World-System Mobility and Economic Growth, 1980-2000.” Social Forces 88 no.
3 (2011): 1123-1152.

30
DeReuck, A.V.S. and Julie Knight, eds. Conflict in Society. London: K. & A. Churchill, Ltd.,
1966.

Hafner-Burton, Emilie M. et al. “Network Analysis for International Relations.” International

Organization 63 (2009): 559-592,

Kali, Raji, and Javier Reyes. “The Architecture of Globalization: a Network Approach to
International Economic Integration,” Journal of International Business Studies 38 (2007): 595-
620.

Kick, Edward , and Byron Davis. “World-System Structure and Change: An Analysis of Global
Networks and Economic Growth Across Two Time Periods.” American Behavior Scientist 44
no. 10 (2001): 1561-1571.

Kim, Jang Hyun, and George A. Barnett. “A Structural Analysis of International Conflict: From
a Communication Perspective.” International Interactions 33 (2007):135-167.

Knoke, David. Political Networks: The Structural Perspective. New York: Cambridge
University Press, 1990.

Lee, Cheol-Sung, Francois Nielse, and Arthur S. Alderson. “Income Inequality, Global
Economy and the State.” Social Forces 86 no. 1 (2007): 77-112.

Lee, Cheol-Sung. “Income Inequality, Democracy, and Public Sector Size.” American
Sociological Review 70 (2005): 158-181,

Levitsky, Steven and Maria Victoria Murillo. “Variation in Institutional Strength.” Annual
Review of Political Science 12 (2009): 115-33.

Mahoney, James. “Revisiting General Theory in Historical Sociology.” Social Forces 83 no. 2
(2004): 459-489.

Mahutga, Matthew C. et al. “Within-Country Inequality and the Modern World-System: A

Theoretical Reprise and Empirical First Step.” Journal of World-Systems Research 27 no. 2
(2011): 279-307.

Mahutga, Matthew C. “The Persistence of Structural Inequality? A Network Analysis of

International Trade, 1965 – 2000.” Social Forces 84 no. 4 (2006) 1863-1889.

Newman, M. E. J. “Mixing Patterns in Networks.” Physical Review 67 no. 2 (2003): 026126-2

– 026126-13.

Perc, M. “The Matthew Effect in Empirical Data.” The Journal of the Royal Society Interface
11, no. 98 (2014): 1-15.

31
Prew, Paul. “World-Economy Centrality and Carbon Dioxide Emissions: A New Look at the
Position in the Capitalist World-System and Environmental Pollution.” Journal of World-
Systems Research 16 no. 2 (2010): 162-191.

Puchala, Donald J. “International Transactions and Regional Integration.” International

Organization 24, no. 4 (1970): 732-763.

Snyder, David and Edward Kick. “Structural Position in the World System and Economic
Growth, 1955-1970: A Multiple-Network Analysis of Transnational Interactions.” The American
Journal of Sociology 84 no. 5 (1979): 1096-1126.

Su, Tieting., “Myth and Mystery of Globalization: World Trade Networks in 1928, 1938, 1960,
and 1999.” Review 25 no. 4 (2002): 351-392.

Ward, Michael D. et al. “Network Analysis and Political Science.” Annual Review of Political
Science 14 (2011): 245-64.

Wasserman, Stanley, and Katherine Faust. Social Network Analysis: Methods and Applications.
Cambridge: Cambridge University Press, 1994.

32